scholarly journals Using Ambulatory Syndromic Surveillance Data for Chronic Disease: A BMI Case Study

2015 ◽  
Vol 7 (1) ◽  
Author(s):  
Andrew Walsh
Keyword(s):  
Syndromic Surveillance ◽  
Disease Outbreak ◽  
Surveillance Data ◽  
Outbreak Detection ◽  
Infectious Disease Outbreak ◽  
Disease Outbreak Detection ◽  
Health Domains ◽  
Ambulatory Practice ◽  
Practice Care

Ambulatory practice syndromic surveillance data needs to demonstrate utility beyond infectious disease outbreak detection to warrant integration into existing systems. The nature of ambulatory practice care makes it well suited for monitoring health domains not covered by emergency departments. This project demonstrates collection of height and weight measurements from ambulatory practice syndromic surveillance data. These data are used to calculate patient BMI, an important risk factor for many chronic diseases. This work is presented as a proof-of-principle for applying syndromic surveillance data to additional health domains.

scholarly journals Supervised Learning for Automated Infectious-Disease-Outbreak Detection

2019 ◽  
Vol 11 (1) ◽  
Author(s):  
Stephane Ghozzi ◽  
Benedikt Zacher ◽  
Alexander Ullrich
Keyword(s):  
Infectious Disease ◽  
Machine Learning ◽  
Time Series ◽  
Supervised Learning ◽  
Expert Knowledge ◽  
Disease Outbreak ◽  
Outbreak Detection ◽  
Infectious Disease Outbreak ◽  
Count Time Series ◽  
Disease Outbreak Detection

ObjectiveBy systematically scoring algorithms and integrating outbreak data through statistical learning, evaluate and improve the performance of automated infectious-disease-outbreak detection. The improvements should be directly relevant to the epidemiological practice. A broader objective is to explore the usefulness of machine-learning approaches in epidemiology.IntroductionWithin the traditional surveillance of notifiable infectious diseases in Germany, not only are individual cases reported to the Robert Koch Institute, but also outbreaks themselves are recorded: A label is assigned by epidemiologists to each case, indicating whether it is part of an outbreak and of which. This expert knowledge represents, in the language of machine leaning, a "ground truth" for the algorithmic task of detecting outbreaks from a stream of surveillance data. The integration of this kind of information in the design and evaluation of algorithms is called supervised learning.MethodsReported cases were aggregated weekly and divided into two count time series, one for endemic (not part of an outbreak) and one for epidemic cases. Two new algorithms were developed for the analysis of such time series: farringtonOutbreak is an adaptation of the standard method farringtonFlexible as implemented in the surveillance R package: It trains on endemic case counts but detects anomalies on total case counts. The second algorithm is hmmOutbreak, which is based on a hidden Markov model (HMM): A binary hidden state indicates whether an outbreak was reported in a given week, the transition matrix for this state is learned from the outbreak data and this state is integrated as factor in a generalised linear model of the total case count. An explicit probability of being in a state of outbreak is then computed for each week (one-week ahead) and a signal is generated if it is higher than a user-defined threshold.To evaluate performance, we framed outbreak detection as a simple binary classification problem: Is there an outbreak in a given week, yes or no? Was a signal generated for this week, yes or no? One can thus count, for each time series, the true positives (outbreak data and signals agree), false positives, true negatives and false negatives. From those, classical performance scores can be computed, such as sensitivity, specificity, precision, F-score or area under the ROC curve (AUC).For the evaluation with real-word data we used time series of reported cases of salmonellosis and campylobacteriosis for each of the 412 German counties over 9 years. We also ran simple simulations with different parameter sets, generating count time series and outbreaks with the sim.pointSource function of the surveillance R package.ResultsWe have developed a supervised-learning framework for outbreak detection based on reported infections and outbreaks, proposing two algorithms and an evaluation method. hmmOutbreak performs overall much better than the standard farringtonFlexible, with e.g. a 60% improvement in sensitivity (0.5 compared to 0.3) at a fixed specificity of 0.9. The results were confirmed by simulations. Furthermore, the computation of explicit outbreak probabilities allows a better and clearer interpretation of detection results than the usual testing of the null hypothesis "is endemic".ConclusionsMethods of machine learning can be usefully applied in the context of infectious-disease surveillance. Already a simple HMM shows large improvements and better interpretability: More refined methods, in particular semi-supervised approaches, look thus very promising. The systematic integration of available expert knowledge, in this case the recording of outbreaks, allows an evaluation of algorithmic performance that is of direct relevance for the epidemiological practice, in contrast to the usual intrinsic statistical metrics. Beyond that, this knowledge can be readily used to improve that performance and, in the future, gain insights in outbreak dynamics. Moreover, other types of labels will be similarly integrated in automated surveillance analyses, e.g. user feedback on whether a signal was relevant (reinforcement learning) or messages on specialised internet platforms that were found to be useful warnings of international epidemic events.

scholarly journals Enhancing Epidemic Detection Using Syndromic Surveillance and Early Notification Methods

2018 ◽  
Vol 10 (1) ◽  
Author(s):  
Tippa Wongstitwilairoong ◽  
Saranath Lawpoolsri Niyom ◽  
Ngamphol Soonthornworasiri ◽  
Jariyanart Gaywee ◽  
Jaranit Kaewkungwal
Keyword(s):  
Early Warning ◽  
Predictive Value ◽  
Syndromic Surveillance ◽  
Detection Method ◽  
Disease Outbreaks ◽  
High Sensitivity ◽  
Disease Outbreak ◽  
Surveillance Data ◽  
Outbreak Detection ◽  
Detection Methods

ObjectiveThis paper presents an investigation using early notification methods to enhancing epidemic detection in syndromic surveillance data from royal Thai army in Thailand.IntroductionEarly Notification Detection Systems have taken a critical role in providing early notice of disease outbreaks. To improve the detection methods for disease outbreaks, many detection methods have been created and implemented. However, there is limited information on the effectively of syndromic surveillance in Thailand. Knowing the performance, strengths and weakness of these surveillance systems in providing early warning for outbreaks will increase disease outbreak detection capacity in Thailand.MethodsThis study describes and compares the capabilities of various outbreak detection algorithms using 37,043 unique syndromic daily reports based on medical information from both civilian and military personnel from the Unit Base Surveillance of Royal Thai Army (RTA) along the Thai-Myanmar and Thai-Cambodia boarder areas. Traditional epidemic detection method: mean plus two SD were compared with algorithms for early notification methods and which included regression, regression/EWMA/Poisson, CDC-C1, CDC-C2 and CDC-C3. Early notification and epidemic detection methods were compared according to their ability to generate alert notifications. Sensitivity, specificity, positive predictive value (PPV), negative predictive value and overall accuracy to detect or predict disease outbreaks were estimated.ResultsThis study shows that the preliminary results are promising for epidemic detection by early notification methods in syndromic surveillance in Thailand. The majority of syndromic records were categorized into 12 symptoms. The three most common symptoms were respiratory, fever and gastrointestinal illness (11,501; 9,549 and 4,498 respectively). The results from the early notification systems were analyzed and their performances were compared with traditional epidemic detection method according to their ability to generate early warning alerts for the 3 symptoms. In our study regression/EWMA/Poisson method had higher specificity across the 3 symptoms (94.5%, 94.7% and 95.9% respectively), but generated lower sensitivity (22.6%, 40.4% and 23.1%). CDC-C1, CDC-C2 and CDC-C3 algorithms are easy to understand and are widely used. CDC-C3 had higher sensitivity to detect gradual disease outbreak effects (64.2%, 70.2% and 57.7%), but it is known to produce higher alarm rates/false positive signals.ConclusionsWithin the syndromic surveillance data of RTA, the CDC algorithm is the best chosen to use in the syndromic system due to being easy to understand and implement in a system with high sensitivity. CDC-C2 is the best early notification detection method due to its high sensitivity and PPV. However, CDC-C3 is shows the highest sensitivity, but exhibits the lowest specificity and PPV for all symptoms including a high alarm rates. To be useful, early notification detection methods must have acceptable operating characteristics. Consequently, we should select the most appropriate algorithm method to explain the data well and in order to improve detection of outbreaks. The comparison methods used in this study may be useful for testing other proposed alert threshold methods and may have further applications for other populations and other diseases.References1. Chretien JP, Burkom HS, Sedyaningsih ER, Larasati RP, et al. Syndromic Surveillance: Adapting Innovations to Developing Settings. PLoS Medicine 2008; vol 5: page 1-6.2. Burkom HS, Elbert Y, Magruder SF, Najmi AH, Peter W, Thompson MW. Developments in the roles, features, and evaluation of alerting algorithms for disease outbreak monitoring. Johns Hopkins APL Technical Digest 2008; vol 27: page 313.

scholarly journals Unknown Disease Outbreaks Detection: A Pilot Study on Feature-Based Knowledge Representation and Reasoning Model

2021 ◽  
Vol 9 ◽  
Author(s):  
Rui Feng ◽  
Qiping Hu ◽  
Yingan Jiang
Keyword(s):  
Knowledge Representation ◽  
Syndromic Surveillance ◽  
Empirical Studies ◽  
Medical Knowledge ◽  
Disease Outbreaks ◽  
Early Response ◽  
Disease Outbreak ◽  
Outbreak Detection ◽  
Knowledge Representation And Reasoning ◽  
Disease Outbreak Detection

Background: The outbreak of COVID-19 in 2019 has rapidly swept the world, causing irreparable loss to human beings. The pandemic has shown that there is still a delay in the early response to disease outbreaks and needs a method for unknown disease outbreak detection. The study's objective is to establish a new medical knowledge representation and reasoning model, and use the model to explore the feasibility of unknown disease outbreak detection.Methods: The study defined abnormal values with diagnostic significances from clinical data as the Features, and defined the Features as the antecedents of inference rules to match with knowledge bases, achieved in detecting known or emerging infectious disease outbreaks. Meanwhile, the study built a syndromic surveillance base to capture the target cases' Features to improve the reliability and fault-tolerant ability of the system.Results: The study combined the method with Severe Acute Respiratory Syndrome (SARS), Middle East Respiratory Syndrome (MERS), and early COVID-19 outbreaks as empirical studies. The results showed that with suitable surveillance guidelines, the method proposed in this study was capable to detect outbreaks of SARS, MERS, and early COVID-19 pandemics. The quick matching accuracies of confirmed infection cases were 89.1, 26.3–98%, and 82%, and the syndromic surveillance base would capture the Features of the remaining cases to ensure the overall detection accuracies. Based on the early COVID-19 data in Wuhan, this study estimated that the median time of the early COVID-19 cases from illness onset to local authorities' responses could be reduced to 7.0–10.0 days.Conclusions: This study offers a new solution to transfer traditional medical knowledge into structured data and form diagnosis rules, enables the representation of doctors' logistic thinking and the knowledge transmission among different users. The results of empirical studies demonstrate that by constantly inputting medical knowledge into the system, the proposed method will be capable to detect unknown diseases from existing ones and perform an early response to the initial outbreaks.

scholarly journals Automated, Laboratory-based System Using the Internet for Disease Outbreak Detection, the Netherlands

2003 ◽  
Vol 9 (9) ◽  
pp. 1046-1052 ◽  
Author(s):  
Marc-Alain Widdowson ◽  
Arnold Bosman ◽  
Edward van Straten ◽  
Mark Tinga ◽  
Sandra Chaves ◽  
...  
Keyword(s):  
The Netherlands ◽  
Disease Outbreak ◽  
Outbreak Detection ◽  
The Internet ◽  
Disease Outbreak Detection
Sign in / Sign up

Export Citation Format

Share Document